mroctopus / Dockerfile
Ewan
Replace madmom with librosa for drum onset detection
00c09df
FROM node:20-slim AS frontend
WORKDIR /build
COPY app/package.json app/package-lock.json* ./
RUN npm ci
COPY app/ .
RUN npm run build
# --- Python backend ---
FROM python:3.12-slim
# System deps: ffmpeg for audio processing, yt-dlp needs it too
RUN apt-get update && apt-get install -y --no-install-recommends \
ffmpeg \
&& rm -rf /var/lib/apt/lists/*
WORKDIR /app
# Install Python dependencies
# basic-pitch pulls in tensorflow on Linux, but we only use ONNX runtime.
# Install it with --no-deps and manually specify what we need.
COPY api/requirements.txt /app/api/requirements.txt
RUN pip install --no-cache-dir \
fastapi uvicorn[standard] python-multipart \
onnxruntime pretty_midi librosa scipy numpy "setuptools<81" \
yt-dlp mir-eval resampy scikit-learn && \
pip install --no-cache-dir --no-deps basic-pitch
# Install Demucs for full-song source separation (CPU-only PyTorch)
# Must install torch+torchaudio from CPU index FIRST, then demucs,
# otherwise demucs pulls CUDA torchaudio from PyPI (2GB+).
RUN pip install --no-cache-dir \
torch torchaudio --index-url https://download.pytorch.org/whl/cpu && \
pip install --no-cache-dir torchcodec demucs
# Copy application code
COPY transcriber/ /app/transcriber/
COPY api/ /app/api/
# Copy built frontend
COPY --from=frontend /build/dist /app/app/dist
ENV PORT=7860
EXPOSE 7860
CMD ["uvicorn", "api.server:app", "--host", "0.0.0.0", "--port", "7860"]